Datasets:

kinokokoro
/

ds2_labeled

You need to agree to share your contact information to access this dataset

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Log in or Sign Up to review the conditions and access this dataset content.

DS2 Labeled

This dataset contains the kinokokoro/ds2 audio shards joined with UtterTune pronunciation labels.

Contents

data/train-xxxxx-of-00917.parquet: Parquet shards with columns:
- audio: original audio payload from kinokokoro/ds2
- text: UtterTune-labeled Japanese text with pronunciation tags such as <PHON_START> and <PHON_END>
- speaker_id: normalized speaker identifier derived from the source metadata

Notes

The source dataset is kinokokoro/ds2. This labeled derivative keeps the same shard count and row order where labels were available.

Generated locally from the DS2 UtterTune labeling pipeline in shisa-tts / Irodori-TTS.